By Jim Ellis — Tuesday, April 21, 2026
Virginia
Today marks the final voting day for the Virginia redistricting referendum. The voters are deciding whether to change the current Old Dominion congressional map from a 6 Democrat/5 Republican split into a gerrymandered plan designed to elect 10 Democrats and one Republican. The measure will likely pass.
The Republicans surprisingly haven’t put up much of a fight considering a four-seat loss in Virginia would substantially decrease the chances of holding their slim five-seat House majority. Democrats, on the other hand, have dominated the airwaves with their ads supporting a “yes” vote on the referendum (see video at top).
The polling suggests the outcome will be a bit closer than the advertising medium suggests (State Navigate; Aril 10-13; 707 likely Virginia voters; 51-45 percent Yes; Washington Post/George Mason University; March 26-31; 1,101 likely Virginia voters; 52-47 percent Yes). Spending on this referendum is completely lopsided with an approximate expenditure ratio of 11:1 in the Democrats’ favor.
Republicans are hopeful that the state Supreme Court, which is hearing an appeal regarding a lower court judge’s earlier ruling that the entire referendum process is invalid because the legislature violated their own rules in qualifying the special election vote, will uphold the previous decision and declare the referendum invalid. The justices have yet to render a final decision but allowed the referendum vote to continue.
It is difficult to see a panel of judges overturning a vote of the people assuming the “yes” position is victorious. Therefore, Democrats will likely be successful in getting the designed 10 Virginia seats. Doing so will go a long way toward achieving the party goal of flipping the Republicans’ slim House majority.
Opposite Polls
The polling industry has absorbed some attacks regarding its cumulative accuracy rate over the past few years and recent polls from Alabama and Michigan underscore that polling is not an exact science.
Part of the problem is how survey research is used. Polling should be thought of as a guide monitoring candidate momentum swings rather than an instrument that tallies exact ballot test numbers.
Sometimes we see situations where polls conducted of the same race during relatively the same time period can produce very different ballot test results.
A good example of this is found in recent surveys from the Alabama Senate and Michigan Governor’s campaigns, both in Republican primaries.
In the Michigan Governor’s contest, we see two recent polls yielding virtually opposite ballot test results. During the April 11-13 period, Emerson College released their poll projecting businessman Perry Johnson to be at parity with Congressman John James (R-Farmington Hills), in fact leading by a slim 21 to 20 percent margin. This is the first such poll that finds Rep. Jones, also a two-time statewide Republican nominee, trailing in the 2026 gubernatorial primary.
Conversely, the OnMessage polling firm, surveying for the James campaign, sees their candidate in the top position with a big lead. According to OnMessage, James has a wide 41 to 18 percent lead over Johnson. While the Emerson College poll was conducted during the April 11-13 period, the OnMessage survey was in the field during the stretch between April 15-18.
It is hard to fathom, however, that the race could have changed to such a great degree in this short time period. Therefore, one of these polls, and probably the Emerson College survey, has badly missed the mark.
In the Alabama Senate Republican primary, we see a similar situation. During the April 11-13 period, Peak Insights released a survey showing Congressman Barry Moore (R-Enterprise) holding a 34-16-12 percent lead over Alabama attorney general Steve Marshall and anti-human trafficking activist Jared Hudson.
Yet, during almost the identical time frame, April 11-14 in this case, the Tarrance Group also surveyed the Republican electorate and found a much closer result with Moore recording only a slight 28-27-24 percent margin over Marshall and Hudson, respectively.
So again, we see two polls commissioned almost simultaneously and sampling within the same macro campaign universe but arriving at very different ballot test results. These results underscore that polling is not exact and that anomalies do frequently happen.
The Alabama and Michigan examples, though relatively extreme, show us that polling again should be looked at simply as a way to judge candidates’ momentum as opposed to precisely pinpointing where they stand in relation to a projected election outcome.